Classification of audio signals on a telephone line

Abstract

This telephone solicitation system provides automatic dialing and call-progress detection. Audio signals on the line are analyzed to recognize network operation tones, noise bursts, and speech. A "sample" is the total number of crossings during a 25 millisecond period. A "window" is a sequence of eight samples analyzed for spectral and time patterns.

Having thus described my invention, what I claim as new and desire to secure by Letters Patent is as follows: 
     
       1. A method of classification of audio signals, including call progress tones and voice, on a telephone line comprising the steps of: sampling an audio signal by counting the number of zero crossings of the signal during a predetermined time period to generate a first value;   using said first value as an index into a first array in which are stored band numbers corresponding to a plurality of possible values to produce a band number corresponding to said first value;   addressing with said band number of second array in which are stored counts corresponding to said band numbers to increment a count corresponding to said band number;   testing said count to determine if a threshold is exceeded to classify the audio signal as a call progress tone; and   examining a number of bands and directions of changes of bands to classify the audio signal as voice.   
     
     
       2. The method of recognizing audio signals as recited in claim 1 comprising the steps of: detecting a change of state of the audio signal to determine when a tone has begun on the telephone line by requiring two successive samples to be above a predetermined noise threshold; and   then performing an analysis on a predetermined number of samples comprising a window for the purpose of classifying the tone.   
     
     
       3. The method of classification of audio signals as reacted in claim 2 wherein the step of performing an analysis comprises the steps of: timing the duration of detected tones; and   comparing the timed durations with possible call states to classify detected tones.   
     
     
       4. The method of classification of audio signals as recited in claim 1 wherein said step of sampling comprises using a predetermined sampling time period and dividing the number of zero crossings counted during that time period by two to derive a number representing an average frequency of the audio signal. 
     
     
       5. The method of classification of audio signals as recited in claim 4 further comprising the steps of: analyzing a predetermined number of successive samples of audio signal, said predetermined number of successive samples defining a window; and   assigning a band number to each of said samples of said window based on said sep of analyzing.   
     
     
       6. The method of classification of audio signals as recited in claim 1 wherein there are a plurality of telephone lines, further comprising the steps of: selecting one of said telephone lines;   dialing an outgoing call on the selected telephone line; and   monitoring the audio signal on said selected telephone line.   
     
     
       7. The method of recognizing audio signals as recited in claim 6 comprising the steps of: detecting a change of state of the audio signal to determine when a tone has begun on the telephone line by requiring two successive samples to be above a predetermined noise threshold; and   then performing an analysis on a predetermined number of samples comprising a window of the purpose of classifying the tone.   
     
     
       8. A method of classification of an audio signal on a telephone line comprising the steps of: analyzing a predetermined number of successive samples of the audio signal, said predetermined number of successive samples defining a window;   assigning a band number to each of said samples of said window based on said step of analyzing; and   using said band numbers to classify said window into one of several possible results including an answer, a telephone call progress signal, a computer modem or facsimile tone, or noise.   
     
     
       9. The method of classification of an audio signal as recited in claim 8 wherein said step of analyzing comprises determining an average frequency and regularity of the audio signal and assigning a band number to each of said samples and the step of classifying comprises the steps of: using the band number to address an array of band values; and   processing an addressed band value of said array to identify the audio signal.   
     
     
       10. The method of analyzing an audio signal as recited in claim 9 further comprising the steps of: testing said successive samples to determine if the band numbers assigned to said samples are the same or are adjacent; and   if a predetermined number of said band numbers are the same or adjacent, returning a classification for the window of tone.   
     
     
       11. The method of analyzing an audio signal as recited in claim 10 wherein if a predetermined number of said band numbers are not the same or adjacent, then testing the randomness of said audio signal to detect a voice. 
     
     
       12. The method of analyzing an audio signal as recited in claim 9 wherein the step of analyzing further comprises the steps of: detecting a change in direction of frequency of successive samples;   when two successive changes in direction have been detected, accepting the detected change as valid; and   performing an immediate analysis of the window upon accepting a detected change of direction.   
     
     
       13. A method of timing the progress of tones on a telephone line to classify the outcome of a call from a range of possibilities including answer, busy, ringing or noise comprising the steps of: analyzing all possible progress of tones on the telephone line as a hierarchy with a large number of results possible at the start of a call but a greatly reduced number of results after only one or two instances of tone/quiet times;   determining from a duration of said tone/quiet times a progress analysis path to follow in said hierarchy; and   returning within a predetermined period of time a classification of the outcome of the call, otherwise returning a timeout indication.   
     
     
       14. The method of timing the progress of tones recited in claim 13 further comprising the steps of: analyzing the frequency of detected tones; and   detecting changes of frequencies to signal a test for possible voice interruptions.   
     
     
       15. The method of timing the progress of tones recited in claim 14 wherein the step of analyzing comprises the steps of: analyzing a predetermined number of successive samples of the audio signal, said predetermined number of successive samples defining a window;   assigning a band number to each of said samples of said window based on step of analyzing; and   using said band numbers to classify said window into one of several possible results including, an answer, a telephone call progress signal, computer modem or facsimile tone or noise.   
     
     
       16. The method of timing the progress of tones recited in claim 15 wherein said step of analyzing comprises determining an average frequency and regularity of the audio signal and assigning a band number to each of said samples and the step of classifying comprises the steps of: using the band number to address an array of band values; and   processing an addressed band value of said array to identify the audio signal.   
     
     
       17. The method of timing the progress of tones recited in claim 16 further comprising the steps of: testing said successive samples to determine if the band numbers assigned to said samples are the same or are adjacent; and   if a predetermined number of said band numbers are the same or adjacent, returning a classification for the window of tone.   
     
     
       18. The method of analyzing an audio signal as recited in claim 17 wherein if a predetermined number of said band numbers are not the same or adjacent, then testing the randomness of said audio signal to detect a voice. 
     
     
       19. A method for classifying as an audio signal within a predefined frequency bandwidth comprising the steps of: dividing said predefined frequency bandwidth into a plurality of smaller frequency ranges called bands;   sampling said audio signal and assigning each sample to one of said bands corresponding to its frequency;   analyzing a number of times successive samples change direction from a higher to a lower frequency band or from a lower to a higher frequency band; and   analyzing the width of a frequency range of said samples and classifying said audio signal according to said number times successive samples change direction and said frequency range width.   
     
     
       20. The method for classifying an audio signal as recited in claim 19 wherein when he number of times successive samples change direction exceeds a predetermined number, the step of analyzing the width of a frequency range is performed by subtracting a lowest frequency band from a highest frequency band to which samples are assigned and classification of said audio signal is made as an oscillating tone the average frequency of which oscillates between tow adjacent frequency bands when the width of the frequency range has a band spread of two or less but as an irregular tone the average frequency of which oscillates between two widely separated frequency bands when the width of the frequency range has a band spread of three or more. 
     
     
       21. The method for classifying an audio signal as recited in claim 20 wherein when the number of times successive samples change direction does not exceed said predetermined number, classification of the audio signal is made as an irregular signal with no average frequency and therefore voice.